Skip to Main content Skip to Navigation

Réseaux d'Automates Stochastiques : Analyse transitoire en temps continu et algèbre tensorielle pour une sémantique en temps discret

Abstract : This thesis presents methods and algorithms for the performance evaluation of large state space models described by high-level formalisms. Among the various formalisms commonly used we use Stochastic Automata Networks (SAN) formalism. SAN formalism is characterized by the representation of very large systems by the composition of its subsystems (automata), where these automata interact with each other by synchronizing events or rates and functional probabilities.

The first part of this thesis focuses on the computation of transient performance indices for large models. When we compute transient indices, such as point availability, the uniformisation method is often used. However, the number of iterations (vector-matrix multiplication) can be very large, which is critical for very large models. Stationnarity detection methods can reduce the computational cost by stopping the iterations when the steady state is reached. In this thesis, we propose an adaptation and a comparison of the different stationnarity detection methods when the matrix is stored in a tensor format. The methods are compared using two criteria: number of iterations and results accuracy.

In the second part, we present the SAN formalism for discrete time models. The SAN formalism formal definition presented in this thesis allows us to define the semantic of discrete time models that we wish to explore. We define a new tensor algebra (called Complex Tensor Algebra - XTA) capable of expressing this semantic. For that uses, three operators are defined to describe the different behavior of a system, such as simultaneity, competition and choice. Finally, the main contribution of this thesis lies in the definition of a tensor formula (called discrete descriptor) that uses this new algebra to describe a discrete time SAN model in a compact mode. We show that this discrete descriptor can easily generate the Markov chain represented by the SAN model.
Complete list of metadatas
Contributor : Leonardo Brenner <>
Submitted on : Friday, October 16, 2009 - 4:35:37 PM
Last modification on : Thursday, October 24, 2019 - 10:35:22 AM
Document(s) archivé(s) le : Tuesday, October 16, 2012 - 12:25:30 PM


  • HAL Id : tel-00424652, version 1


Leonardo Brenner. Réseaux d'Automates Stochastiques : Analyse transitoire en temps continu et algèbre tensorielle pour une sémantique en temps discret. Modélisation et simulation. Institut National Polytechnique de Grenoble - INPG, 2009. Français. ⟨tel-00424652v1⟩



Record views


Files downloads